Deep Learning Approach to Identify a Person using GAIT Analysis

This Model introduces a novel deep learning-driven system for human identification through gait analysis—a method centered on recognizing individuals based on their walking style. Unlike facial recognition, which often fails in low- resolution, occluded, or dimly lit environments, gait offers a non-contact and long-distance biometric alternative. Silhouettes are extracted from video sequences, and temporal-spatial characteristics unique to each person’s movement are captured through convolutional neural networks and dynamic temporal modeling. The framework was trained and evaluated using publicly available datasets like PSYMO, showcasing commendable recognition accuracy across varying conditions such as attire changes and shifting viewpoints. Beyond improving surveillance capabilities, this approach is crafted to prioritize ethical boundaries and user privacy.